AI's Impact on Modern Warfare: A Deep Dive

0:00Many people are used to the ways that ChatGPT can help them write a better article or AI can help them make better decisions in small tasks. Today on the podcast we are going to be talking about how chat GPT's technology large language models and how AI in general is being used in the defense industry by the military and by different governments. Today we're going to be talking about a couple big companies that have recently come out with different defense related products in AI using large language models and we're going to talk about some of the implications of those tools. So the first one that I wanted to cover is Palantir because you may have heard it in the news lately.

0:39Recently, it's backed by Peter Thiel, I believe. And recently, they did a demo showing one of their AI large language model kind of based AI tools that aggregates data and helps to make military decisions and operations. And when they did that, their stock, which they're a publicly traded company, because I believe that they're kind of like a data analytics software vendor. And their stock price went up 20%. So it was this big thing. And apparently they're having more demand than ever for their current software platform that they have released, which is called AIP. And that stands for Artificial Intelligence Platform.

1:23That sounds really generic. So I think I would have picked a better brand name. but regardless obviously they're the ones that are incredibly successful and doing well on this so what do I know but in any case essentially their company's tool can be used by militaries to tap the same AI that that's chat GPT is built on and that is to help in battlefield intelligence and decision making so they had a recent demo video that kind of showed how their platform worked and I watched it it was really fascinating but essentially what they're doing is they're analyzing intel on enemy targets they identify potential you know hostile situations or things like that then they'll go and they'll ask the ai to propose a battle plan and then they'll have the ai send the battle plan over to the commanding officers to kind of work on the execution and it's honestly really impressive how this is done there's recently another tool and company that has come out with a very similar tool called donovan which i'll cover in a second but essentially just to give you an idea of like how this works really because um when i first heard about it it was kind of hard to visualize if you can kind of imagine it they have kind of like a command portal with a chat model on the one side and on the other side is like a live satellite view of a map and on that map you can have different data points now it's important to note that this whole system would be tied directly into all like it would have like access to all the confidential classified information of um of a military essentially or any sort of group so it would know what the locations of your different teams your different artillery um it would know essentially it's kind of like it's interesting to think about this like a company but like you know someone like walmart they have like point of sale systems um and they have different systems that track their inventory how many bottles of ketchup are on one shelf.

3:18So military's got the same thing, except it's like how much ammo is in different locations, how many missiles does this team have, how many tanks are over here. So they have the same kind of systems that track and all of that data is put into this tool that has a large language model. And what's really interesting about this is it is based off of your military doctrine. So I'm assuming they're probably gonna have exclusive contracts with the US government. That would be interesting if they were doing foreign governments and then, you know, two governments went to war or had some sort of squirmishes and they were both using the same sort of AI-based technology to fight each other.

3:55So I'm assuming there's going to be some sort of exclusive contracts on this kind of thing. Obviously, the defense industry spends very heavily on this, right? Like if the U.S. government could get a 20 % advantage over an adversary by using an AI tool like this, I think they'd pay a lot for it. especially if it can help decrease mortalities and whatnot. So in any case, this thing plugs into all of your data and it is integrated with more than just your data of where things are but the ability to send out things like autonomous drones. So in the example they gave, essentially the person said, hey, scan the map and let me know where my adversaries are on the map.

4:34So it has like a scan of the map and based off of the latest Intel data, which by the way they said they could anytime someone uploads a new PDF or a new image to like a file that's that they're able to scan it it's updating in real time so it knows the position of everything and it has you know GPS searching through everything so they have they they then have it say like you know alert me of any anomalies that you see on this map and it will say okay well an anomaly is that you know this tank here has moved forward this amount of space that you know know we weren't expecting that blah blah blah and then you can say okay give me three um possible like uh pieces of action that i could do to help mitigate this risk or you know engage with this adversary and it would say okay so you know over here you have like the 159th artillery you could fire that at it or over here you have like an airbase you could fly airplane over and you know try to shoot it or there's this team omega here that has a bunch of anti-tank javlin missiles you could send them up so it has a bunch of different options and then you could say okay send these options up to command control it has the whole like military brief mapped out and sent over to your commanding officer and then you can also say all right I'm interested in this I would like to at the same time gather more intel so send out a what's the best way to gather intel on this they would say okay well you could you know you could fly an airplane over it you could use this type of a team to go and spy on it or you could use like some sort of drone and you say okay use the one that requires the least amount of like humans so they're like all right we select the drone it would send in a drone to go fly over take pictures it would bring the pictures back to you immediately on the platform where you can analyze and also send to your higher ups to justify or decide if it is an actual threat or not really really powerful technology and very very interesting um palantir when they did this demo um which was it's you know very it looks like a very sophisticated piece of software um you know showing live footage from airplanes of tanks and all sorts of things uh it obviously got a lot of interest from a lot of different people and um their company has a lot of interest in that so really interesting to see kind of how the defense industry is using this it's like this large language model and decision making based off of it is really powerful.

6:56Something I think that they're doing that is really useful is the fact that this isn't just like a black box of, you know, you ask like a chat GPT, for example, based off of these things, what should I do? It's saying based off of these, these things, and our current military doctrine, our, you know, engagement strategies that are x, y, and z, and all of this data that we have of how we typically respond to these kind of things, what would you recommend the best thing to do that would have the least amount of casualties that would, you know, neutralize the threat the fastest you can have all your different stipulations in there and it's based off of that and unlike chai chipputi that is a black box it will show you in the back end you can kind of open up the hood and look at the back end and see which ai model because you can actually put different types of ais into here merging multiple different ais together to do these ais that are trained for specific parts of these tasks but you can actually open up the hood and look in the background and see which AI model made which decisions, which decisions had to go to actual people, what, why, like which intel people were in the loop of, and which decisions were approved or suggested by people.

8:06So it's really transparent on, you know, it's not just like an AI is all of a sudden taking over the military and making all the decisions. It's really, it has a very clear mapped out, which I think is really important for transparency, especially in something as touchy as military operations or as critical, I would say. And so it really maps out who's responsible for what, who made what decisions and what the AI was doing. And I think it's interesting because it's not just saying like, you know, neutralize the threat. It's saying, give me some options. It creates options and it, you know, it gives all the data about the options.

8:42It's like, you know, if you go and engage with this based off of, you know, the artillery here, it's, you know, 40 kilometers away, you have an 89 % success rate of hitting it, the fuel level on the artillery is at like 14%, the personnel required, there's eight people required to fire that, it's going to take about 18 minutes to get that system up and going, these are your assets that you have there, so it really gives you a lot of different data. And with all of these options that it lays out, it gives you all the data of why it picked those options. And let's, you know, obviously an actual person make the call on that.

9:19So this is not AI that is running and doing the whole thing. So I think it's really impressive technology. All obviously very scary thinking about AI being incorporated into the military, especially because when I look at this, and I see a company coming out of America, I'm living in America, and I'm like oh great this is more defense but i think the scary thing is obviously china russia all these other countries um that could potentially be adversaries for me or for other countries around the world um will be having similar advanced ai and it almost becomes like an ai arms race where the most ai advanced country will win and i think this isn't a this isn't a surprise like countries know this i believe vladimir putin recently said that ai will be like a major determining factor between countries success.

10:07Obviously, America is putting a lot of focus on AI, you hear a lot of a lot of talk out of China about AI and kind of what they're doing there. So really interesting to see how this is going to be integrated into military. And that is what Palantir is currently working on. Now, in addition to kind of what Palantir is going on is doing, I did want to just highlight really quick another company called Donovan that recently launched from scale. So scale as a software company and I also will note that these softwares are definitely you know being used in the military but they're also being used in other areas such as crisis management for companies as well so they have like slightly changed versions of this where you essentially hook this into all of your company data and if you know you're in a area and a hurricane is coming and you have like 10 stores down there it can also help you to mitigate crises down there make plans to evacuate, to lock down things, you know, what stock to move, maybe that's highly valuable, and a whole bunch of different decision making things for people that, for like companies as well.

11:12So these, these softwares kind of go into both of those. But I think right now, a lot of the hype and things that we're hearing people talk about is coming in the defense area. So I will say with Donovan, they recently, you know, they kind of tell on their website, the fact that the US Defense Department used them to come in and kind of evaluate the current AI battlefield systems that they have to make sure that they were up to, I guess, up to par and whatnot. And they also have a lot of, you know, data pulled from the US Army, the Defense Innovation Unit, the CDAO, the AFRL, and a bunch of other different organizations, including the Air Force to like, essentially create this AI model that is really good at this.

11:59And I'm assuming they're not, you know, I'm assuming the U.S. government giving out this type of data has exclusive contracts with them. So they're not going to go take U.S. government data on how we fight and go, you know, sell that to Russians or whatnot. But definitely is a highly contentious area that I'm sure people will be watching very closely. again with Donovan just like Palantir when you use the software it's the same kind of idea where it opens up a it'll open up a portal with a chat on one side and kind of a kind of like a map on the other side in Donovan's demo they showed they asked essentially like hey in the last 48 hours what were all the ship movements of adversarial navy vessels in you know this Asia-Pacific region it highlighted them on the map it said tell me if there's any anomalies in this it highlighted one that was off course from what it was supposed to be doing they were able to then go and get a flyby drone had they had help making the decision on giving the flyby drone to go and get more images when they did the image scan they saw that there was a higher level of radiation around the vehicle and what was really interesting that Donovan did in the demo it was beyond just like the documents that they had, they kind of branched out and said, is there anything in recent like publicity from that country related to higher levels of radiation and advancements in X, Y, and Z military weapons that would, you know, that would explain what this is.

13:30And it actually found a report and then, you know, the report was in Chinese. And so then it had it translate from Chinese, translate the report into English, and they were able to package it and summarize it. So they're like, okay, summarize this entire report. And based off of the other things we've talked about, please put this into like a report I can send to my higher up commanding officer. And so I really see, anyways, they went through this whole exercise. And in the course of like eight minutes, they went from, you know, find me an anomaly, gathered all the data they needed, um, and figured out kind of what was going on and sent a report that they could then go and take a course of action, which doing like a flyby on an airplane, um, to, you know, check what was going on with this vessel.

14:08So in the course of minutes, you can have these AIs do really complex tasks that would usually take military analysts days to do. And I think this is really going to accelerate the rate of protection in the defense industry. As wars go on, it's going to be scary to see the speed that AI models are able to do things. um if what I would be worried about to be honest in like a forward-thinking manner is obviously when you have to get like human approvals and reviews of different things it takes longer for a human to look over all the data make sure everything checks out and then be like um you know decide like okay it's 18 minutes to do this course of action and 10 minutes to do that course of action um so we're going to go with this one or that one what I would be worried about is different militaries around the world that are deciding, hey, we're just going to, you know, to make this thing faster, we're just going to say, remove human intervention, unless it has like a probability of more than 20 casualties.

15:04And just like, like, they're going to make decisions like that, that will be a variable in some of these AI models, where the AI is, is going to be making some decisions of life or death. And in an in the effort to speed up response times, sometimes people will be removed from the equation. That is what I think I'm most concerned about is not necessarily, I think it's amazing, especially for the military defending myself to have tools like this that can help it make better decisions faster. But what scares me is the probability that these eventually are the human element will be removed from this.

15:39And the fact that an AI will be rapidly iterating and making decisions that will affect human life. So I think it's going to be an area to definitely focus on deeply. Do I think it's a bad area we shouldn't be focusing on? It's hard to say because in the concept of national defense, if we are not developing these, China, Russia, other geopolitical adversaries will. And so I definitely think we 100 % need these tools. you know it's like the atomic bomb i wish it didn't exist but because it does and because that technology is there impossible like it definitely needs to be developed uh where where you are in order to be a defense so it's gonna be really interesting i think ai will be heavily integrated in the defense industry in the future going forward uh and i'm sure it already is now to some degree but um in fact i was recently talking to a friend who um told me that while he was in South Africa.

16:37It was the one in the military a couple years ago and gathering intelligence is one of his first introductions to actually not South Africa. It was like the Congo. It was one of his first introductions to AI. They had these really advanced AI systems that were tracking people and facial recognition and very interesting to hear about. You know, he was like, why are we developing this in the Congo instead of the US and other questions like that. There's all sorts of ethical questions. Of course, we probably test all this AI stuff in other countries, not farther away from our, you know, US citizens that might complain or whatnot.

17:12So it's really interesting. But definitely, the AI is definitely already integrated into the defense system. And this is kind of going to be the next step where it helps to arrange everything and make battle plans and suggestions on the battlefield. So definitely an area we're going to be watching in the future. So stay tuned as we continue to discuss this topic.